This paper explores how the emergence of generative artificial intelligence is reshaping the information environment in capital markets. Leveraging an unexpected ban on ChatGPT in Italy, we examine its impact on the information processing capabilities of market participants. We employ metrics for AI-generated text detection to show that the ban coincides with decreased AI usage by domestic financial analysts and fewer earnings forecasts issued relative to foreign analysts covering the same firm. The negative effects are more pronounced among analysts whose pre-ban reports are more consistent with AI use or analysts with a technical background. The ban also diminishes forecast accuracy, increases reliance on industry-specific information, and reduces information efficiency. Furthermore, investor reactions to earnings announcements become more pronounced, and bid–ask spreads widen, reflecting lower market efficiency.

Investors have a finite capacity to organize all information they receive from financial disclosures. Under rational inattention, we show that investor processing capacity affects the probability of disclosure. Our main result is that the likelihood of disclosure is inverse-U shaped in investor attention. For low levels of attention, more attention facilitates communication and increases disclosure; for high levels of attention, more attention better identifies, and therefore deters, unfavorable voluntary disclosures. We provide empirical evidence that the relationship between investor attention and management forecast follows the predictions of the theory, using institutional ownership as a proxy for investor attention as well as exogenous shocks to fund manager distraction.